| Summary: | 博士 === 國立臺灣大學 === 流行病學研究所 === 88 === ABSTRACT
Objects. Most cancer mapping methodologies are county-based. However, the trend surface analysis point-based has the smoothing surface but suffers from the edge effect. two new modified trend surface analyses therefore are proposed to diminished the effect. Finally, the author also demonstrates it’s application in spatio-temporal characterization.
Methods. The author compares six methods (SMR, SMREB, SMRadj, PTS and the newly proposed PPTS, IPTS) by calculating the average variance, square bias, mean square error, percentiles and the power for identifying the high risk location through simulations. In addition to aforementioned new methods of trend surface analysis, contained herein is also a study demonstrating the applicability of the IPTS methodology to a dynamic spatio-temporal trend using the male oral cancer data.
Results. The results show that both PPTS and IPTS can dramatically diminished the edge effect. Especially for the IPTS method , it has the minimum values in average variance, square bias, mean square error and has better percentile estimate. Besides, it also has the shortest median error distance when capturing the exact hot spot and clearer dynamic spatio-temporal trend in application. When only compare SMRadj with SMREB, the simulation result shows SMRadj has smaller average means square error than SMREB.
Conclusion. We suggest the SMRadj is a simple and practice method to map the atlas if the coordinates are not available. Otherwise, the IPTS method should be a better choice when cancer mapping.
|
|---|
